Download Primary-Export-Led Growth: The Evidence of Ghana

JOURNAL OF ECONOMIC DEVELOPMENT
Volume 23, Number 1, June 1998
Primary-Export-Led Growth: The Evidence of Ghana
Attah K. Boame
** 2
This paper investigates empirically the causal relation between GNP growth and export growth
for Ghana from 1960 to 1992. The theoretical framework has been the staple theory of growth
which emphasises the crucial role of primary export(s) in the economic growth process. In particular,
the staple theory stresses the linkages (backward, forward and final demand) of the staple export(s)
with the rest of the economy. The paper performs the Granger (1969) causality test for Ghana.
The results support the primary-export-led growth strategy for Ghana. In other words, the evidence
in favour of export promotion is stronger than previous statistical studies have indicated.
I. Introduction
Exports are supposed to enhance economic growth. There is considerable evidence
supporting the export promotion hypothesis as a development strategy. The “success” stories
of Hong-Kong, Singapore, South Korea and Taiwan need no further elaboration. The basic
hypothesis is that growth in real exports leads to growth in real GNP. Trade theorists pinpoint
a number of factors causing this link. Thus, we may postulate that most developing countries’
growth has been influenced by their dependence on primary commodities export. The relevant
economic theory underlying this primary-export-led growth is the staple theory of growth.
The staple theory emphasizes three kinds of benefits to a trading country - improved utilization
of existing resources, expanded factor endowments and linkage effects.1 For Ghana, it may
be interesting to analyse the extent to which the staple theory of primary-export-led growth
is applicable to the economic growth process. Would the continued reliance on primary exports
lead the country to sustained economic growth (as the fur and wheat trade did in early Canadian
economic growth history)? Or is the evidence strongly against primary exports, implying
the need to diversify, especially into manufactured exports? This paper investigates empirically
the role of primary exports in Ghana’s economic growth process, where economic growth
is defined as a rise in national income or GNP.
** I wish to express my appreciation to Prof. W. Simpson for his constructive comments and advice on the paper;
Prof. N. Cameron for his helpful comments on an earlier draft of the theoretical aspects of the paper; Dr. A.
Sepehri and Dr. L. Brown for some suggestions; Prof. H. Rempel for some reading materials; and Dr. B. Troutt
for editing the paper, all at the Department of Economics-The University of Manitoba. I am also grateful to
an anonymous referee of this journal for useful comments. However, I am responsible for any remaining errors
or omissions.
** Department of Economics, The University of Manitoba, Winnipeg, MB., Canada R3T 5V5.
1. These are essentially the same as the dynamic effects of trade emphasized by the theory of comparative advantage:
trade leads to specialization in production; trade provides a “vent-for-surplus” commodities; trade leads to expansion
of production possibilities through competition, access to new knowledge and technology; and trade provides
imports which may be used as substitutes in domestic production.
175
JOURNAL OF ECONOMIC DEVELOPMENT
Section II briefly presents the staple theory of primary-export-led growth as the framework
for the analysis. Section III presents an overview of the composition and role of primary
exports in the Ghanaian economy. The presentation attempts to link the role of the primary
exports to the staple theory within the economic growth process of Ghana. We also compare
Ghana’s comparative advantage in primary commodities with other countries which have
successfully accomplished primary-export-led growth. Section IV deals with the econometric
specification based on the Granger (1969) causality test as the main tool of analysis. Finally,
Section V presents the estimations, analysis and conclusion of the paper.
II. The Staple Theory of Primary-Export-Led Growth
In order to integrate both historical and theoretical aspects of the export promotion
hypothesis, we will adopt the staple theory of primary-export-led growth as the framework
of analysis. 2 In the discussion of the staple theory, the literature of formal mathematical
and geometrical growth models is not emphasised because these are unable to handle conveniently
the number of variables required to explain the complex phenomenon of export-led growth
(Roemer (1970)). The staple approach to the study of economic history is primarily a Canadian
innovation, the leading innovator being Harold Innis in his pioneering historical studies, notably
of the cod fisheries and the fur trade (Grant and Watkins (1993)).
A staple is a primary product that faces a large and growing demand in world markets,
does not require elaborate processing, and has a high enough value-to-weight (volume) ratio
to bear transportation costs. In other words, it is a profitable primary commodity export.
Such an export industry may be established as a consequence of recent discovery, increased
demand, cost-reducing technological change, or any combination of these. Staple production
necessarily uses natural resources intensively.
The basic assumption of the staple theory is that staple exports are the leading sector
of the economy. That is, the export sector acts as a key propulsive sector, propelling the
rest of the economy forward. The limited - at first possibly nonexistent - domestic market,
and the factor proportions - an abundance of land relative to labour and capital - create a
comparative advantage in resource-intensive exports, or staples. Economic development will
involve diversification around this export base. The central concept of the staple theory, thus,
is the impact of export activity on domestic economy and society. To recap, the likely growth
path of a staple economy may be the following. Growth is initiated by an increase in demand
for a staple export.3 If the spread effects are potent, as the export sector grows so too will
the domestic sectors. This will lead to increasing demand for factors such as capital and
labour. Domestic slack, if it exists at all, will be quickly absorbed, and the continuation
of growth will depend on the ability to import scarce factors. If the supply of foreign factors
is elastic, the customary tendency for the expansion of one sector - in this case exports to affect domestic sectors adversely by driving up factor prices is mitigated. This explains
2. The presentation here is based on Grant and Watkins (1993) and Roemer (1970).
3. Corden (1971) analyses a “supply-motored” version of the staple theory that stresses growth in factor supplies
and productivity.
176
Primary-Export-Led Growth: The Evidence of Ghana
the very strong booms that are a feature of growth in staple economies (Grant and Watkins
(1993)). We can state the staple theory in the form of a disaggregated multiplier-accelerator
mechanism. The domestic investment resulting from the increased activity of the export sector
can be disaggregated into three linkage effects: backward, forward and final demand linkage.
Backward linkage measures the induced investment in the home-production of inputs,
including capital goods, for the expanding export sector. The most important example of
backward linkage is the building of necessary infrastructure to enable exportation. Facilities
established to serve the staple industry, such as roads, electric and water supplies, will lower
costs in other potential industries and may stimulate further investment.
Forward linkage measures the induced investment in industries which use the export
industry’s output as an input. The most obvious and typically most important is the further
processing of the output from the export sector. The output from the processing industries
may be used domestically or exported. The processed exports increase the value added from
the export sector in foreign markets.
Final demand linkage measures the induced investment in domestic industries producing
inputs (both consumer and intermediate as well as services) for the export sector. Its prime
determinant is the size of the domestic market, which is in turn dependent on the level of
income - aggregate and average - and its distribution. Final demand linkage will tend to
be higher, the higher the level of income and the more equal its distribution.
The staple theory, although very illuminating, is beset by certain drawbacks which include,
among others, (i) staple exporters (specifically, those exercising political control) will develop
an inhibiting “export mentality.” This may result in an overconcentration of resources in the
export sector and a reluctance to promote domestic development; and (ii) the economy may
get caught in a “staple trap” which may involve specialization in the wrong kind of staple.
Sustained growth requires the capacity to shift attention to new foreign or domestic markets.
This necessitates transformation of the institutions and values as the economy grows. In the
absence of such transformation, the growth process propelled by the staple export(s) is likely
to be halted. The next section briefly presents the composition and role of Ghana’s export
sector as well as linkages with the rest of the Ghanaian economy. We also compare Ghana’s
comparative advantage in primary commodities with other countries.
III. The Composition and Role of the Export Sector in the Ghanaian Economy
The presentation here focuses on data from 1960 to the present, based on Huq (1989),
and is brief due to space limitation. Export proceeds are the main source of foreign exchange
needed for the importation of intermediate and capital goods. The contribution of exports
to the economy of Ghana has been substantial. Ghana has traditionally depended on export
revenue as a major source of government revenue (this is a fiscal linkage). Exports comprised
as much as a quarter of GDP during the immediate pre- and post-independence periods.
Export duties comprised, on the average, between 10.2 and 35.3% of central government revenue
from 1975 to 1990 (Kapur et al., (1991)).
A common characteristic of developing countries’ exports is the dependence on one or
a few primary commodities. Ghana’s exports comprise basically primary products, notably
177
JOURNAL OF ECONOMIC DEVELOPMENT
cocoa, minerals (gold, diamonds, bauxite and manganese), and timber products. Since the
mid-1960s, some cocoa beans have been processed into cocoa paste and butter so as to increase
their value for export. In 1965, the quantity of cocoa exported reached its peak of 502,000
tonnes. The cocoa group contributed, on average, 59% of the total value of exports in the
1960s; this share rose to 72% by 1970, and by 1980, cocoa and cocoa derivatives accounted
for about two-thirds of total export earnings. Timber (both logs and sawn) contributed 14%
of total exports in 1960, but dropped to 3.3% in 1980. The share of gold was 10% in 1960,
but contributed only 6% in 1970. In 1980, due to an increase in the world price of gold,
its share rose to 17% of total export earnings. In value terms, the shares of the other minerals
- bauxite, manganese and diamonds - have not been very significant.
1. Linkages
The backward linkage seems to be the most developed among the three linkage effects
of Ghana’s export sector. The network of constructed roads has been expanded since 1900,
and consists of primary, secondary, feeder and town roads. Feeder roads connect the main
roads with rural farming communities. These feeder roads facilitate the transport of foodstuffs
and other raw materials to urban centres. The impact of the railway system on the Ghanaian
economy has been substantial. Gold, manganese and bauxite mining all benefit tremendously
from the railway network. The cocoa, timber and kola-nut industries are also major beneficiaries.
Water transportation, especially on Volta Lake and the ocean, have been developed to meet
the country’s external trade requirements. The Takoradi harbour, which was opened in 1928
and expanded in 1953, is the main exporting port. It handles bulky exports like gold, manganese,
bauxite, timber and cocoa. The Tema harbour, opened in 1962 as part of the Volta River
Project, principally handles imports - basically imported capital goods for the manufacturing
sector.
Electricity supply has also been developed to aid the growth of the economy. The
Volta River Authority (VRA) was established in 1961, with the main function of constructing
a dam at Akosombo on the Volta River for the generation and distribution of power for industrial
and domestic use.
Institutional facilities (e.g., basic research, banking and insurance, etc.) have also been
established. The aim has been to facilitate investment in the productive sectors of the economy.
The Cocoa Research Institute at Tafo undertakes research pertaining to the cocoa industry
with the aim of improving the efficiency and output of the industry. The Cocoa Marketing
Board (CMB) was set up in 1947 to export cocoa. The Bast Fibres Board was also established
in 1970 to produce the needed cocoa bags. Agricultural extension services in the country
started around 1900, with travelling instructors who advised on cultivation of cash crops (mainly
cocoa) and supplied inputs to farmers. Several banking institutions (central, commercial and
rural) operate in the country. The Rural banks serve as catalyst institutions in the economic
development efforts of the rural population.
The forward linkage is the least of the linkage effects of the export sector. It is rather
unfortunate that much of the output of the primary sector is exported in raw rather than processed
form. The only processed goods for export are cocoa butter/paste and plywood/veneer, which
178
Primary-Export-Led Growth: The Evidence of Ghana
form only a small proportion of exports in terms of both value and volume.
The final demand linkage is relatively more developed than the forward linkage. It
comprises the various manufacturing industries and social services which provide inputs and
support for the export sector. As Ghana developed, investment in education and health (using
revenue from the export sector) was undertaken to buttress the growth process. In 1965,
government spent over 5% of GDP on education. Total government expenditures on health
were 1.6% and 1.0% of GDP for 1965 and 1980, respectively.
2. Export Policies
Ghana’s declared goal has always been to increase the volume of exports. Various
policies have been implemented to achieve this objective. For instance, in 1977, the existing
export bonus for non-traditional exports was raised from 20 to 30%, and for traditional exports
(excluding cocoa), a new export bonus of 20% was instituted. Procedures relating to exportation
have been streamlined to reduce inefficient bureaucratic practices. There have also been
attempts at diversification (e.g., processing of timber into veneer and plywood, cocoa into
cocoa butter/paste) in order to capture higher prices in external markets.
In 1979, to encourage timber exports, the government abolished the export duty on timber
and increased the value of import licenses granted to the industry. In 1983, the government
attempted to encourage exports by introducing a highly attractive export bonus scheme in
the annual budget. Under the Economic Recovery Programme (ERP) of 1983, a flexible
exchange rate policy was instituted and has contributed to an expansion in the volume of
exports. In response, export volumes increased at an average annual rate of 10% between
1983 and 1990 (Kapur et al., (1991)). We discuss next Ghana’s comparative advantage in
primary commodities.
3. Ghana’s Comparative Advantage in Primary Exports
This section compares Ghana’s comparative advantage in primary commodities with
Indonesia, Malaysia and Thailand which have successfully accomplished primary-export-led
growth. These countries achieved an impressive expansion of GDP per person of 4 % annually
between 1965 and 1988. The share of primary goods in merchandise exports are 53% for
Indonesia, 39% for Malaysia and 34% for Thailand, and these countries remain important
exporters of primary products (Snodgrass et al., (1995)). They have all gained from their
comparative advantage in primary products. Malaysia, for instance, enjoys a rich resource
base with a relatively small population, and its development strategy has been based on its
comparative advantage in primary products: rubber, palm oil, wood and petroleum. As a
consequence of its investments in primary and manufactured exports, Malaysia sustained rapid
economic growth, over 6 percent a year, from 1965 to 1988 (Snodgrass et al. (1995)).
In order to analyse the comparative advantage of Ghana, Indonesia, Malaysia and Thailand
in primary commodities, we will use the factor endowment and the revealed comparative
advantage indicators. According to the Heckscher-Ohlin model, the source of comparative
advantage lies in differences in factor endowment. The weight carried by different factors
179
JOURNAL OF ECONOMIC DEVELOPMENT
may, however, vary from country to country and from sector to sector. Differences in natural
endowment (land, labour and natural resources) have been taken to be sources of comparative
advantage in the case of raw-materials and agricultural products. Capital (physical and human),
technology, scale economies, research and development have similarly been given high weight
in manufactured products. Factor endowment for an economy can be measured in two ways:
absolute factor endowment and relative factor endowment. We will use Leamer’s (1984) resource
abundance ratio to measure the absolute factor endowment in our analysis. This is defined as:
where
denotes endowment of resource
of resource
,
is GNP of country
in a given country
and
,
is world’s endowment
is world GNP.
The country is posited to be abundantly supplied in factor
compared with other resources
if an endowment share (
) exceeds GNP share (
), that is if
. The absolute resource endowment ratios for land and labour are presented in Table
1. We see that the absolute resource endowment ratio was greater than unity from 1970
to 1990 for Ghana, Malaysia, Thailand and for Indonesia from 1975 to 1990. The results
indicate that Ghana, Indonesia, Malaysia and Thailand have comparative advantage in land
and labour.
Table 1
FACTOR
LAND
LABOUR
YEAR
1965
1970
1975
1980
1985
1990
1965
1970
1975
1980
1985
1990
Land and Labour Abundance Ratios
GHANA
0.1303
1.9320
1.8424
4.4077
4.7411
6.1986
0.1657
2.4397
2.2937
5.3318
6.0299
8.2028
INDONESIA
0.2565
0.3681
2.2860
2.1534
1.9273
2.7725
0.4965
0.7436
4.7293
4.4715
4.1072
6.1242
MALAYSIA
0.1443
1.8527
1.3253
1.1913
1.0364
1.2395
0.1242
1.7280
1.3355
1.2941
1.1694
1.4664
THAILAND
0.1651
1.4432
1.2947
1.3766
1.1967
1.0274
0.4485
4.0727
3.7512
4.2488
3.7750
3.2844
Note: Data on land and labour are taken from FAO Production Yearbook Vol. 30 (1976) and Vol. 45 (1991).
A rigorous analysis of the patterns of comparative advantage, however, requires cost
comparisons across countries. 4 But the data required for such an analysis are rarely available.
4. Domestic resource costs (DRC) and Net social profitability (NSP) (see Bruno (1967)) are the two measures commonly
used in the analysis of production efficiency. The DRC is the ratio of domestic factor costs evaluated at social
opportunity costs to value-added - the domestic resource cost of earning foreign exchange. The NSP is equal
180
Primary-Export-Led Growth: The Evidence of Ghana
The export performance index which reflects all types of costs and non-price factors and based
on the empirical data pertaining to trade flows has often been used as an indicator of
comparative advantage. Since this is revealed by the observed pattern of trade flows, it is
called revealed comparative advantage (RCA) by Balassa (1965). The most popular index
of RCA is the export performance ratio. According to Balassa, the comparative advantage
of a country
in trade of a particular commodity
can be measured by the share of that
commodity in the country’s total exports relative to the commodity’s share in total world exports.
That is:
If this ratio is greater (less) than unity, it is interpreted to mean that the country
has a
comparative advantage (disadvantage) in the export of the product .
The indices of RCA were computed for primary commodities for Ghana and compared
with the other countries as shown in Table 2 below. Again, we see that all these countries
have a comparative advantage in primary commodities from 1966 to 1985. The RCA values
have exceeded unity for Ghana since 1960. The next section discusses the econometric
specification.
Table 2
YEAR
1966
1970
1975
1980
1985
Revealed Comparative Advantage for Primary Commodities
GHANA
2.686
2.907
3.629
6.073
6.500
INDONESIA
2.085
0.931
1.313
2.143
MALAYSIA
2.403
2.336
3.437
2.824
2.600
THAILAND
3.260
2.863
3.520
3.484
1.458
Note: The RCA values for Ghana are computed by the author; the RCA values for Indonesia, Malaysia and Thailand
are taken from Sachdev ((1933) Table 4, p.210). The RCA values for Ghana are 2.970 and 8.300 for 1960
and 1990, respectively.
IV. Econometric Specification
1. Previous Studies of the Export Promotion Hypothesis
Traditionally, the export promotion hypothesis has been tested empirically by regressing
a growth variable on an export variable using OLS techniques. In these studies, the statistical
significance of the export variable coefficient indicates the support for the export promotion
hypothesis. A summary of some of these studies is presented in Table 3. All of the studies
to value-added in world prices less domestic factors evaluated at their social opportunity costs or shadow prices.
These measures can be thought of as indicating the relative comparative advantages of competing activities.
DRC values less than one identify activities in which the country in question appears to have an international
comparative advantage and vice versa.
181
JOURNAL OF ECONOMIC DEVELOPMENT
except those by Jung and Marshall (1985) and Darrett (1987) support the export promotion
hypothesis.
Table 3
Study
Michaely
(1977)
Previous Empirical Studies of Exports and Growth
Data Set
Cross-section
41 countries
Ave of 1950-73
Balassa
Cross-section
(1978)
10 countries
Aves of 1956-67
and 1967-73
Williamson Cross-section
(1978)
22 countries
Ave of 1960-74
Fajana
Time series
(1979)
1954-74
1 country
Tyler
Cross-section
(1981)
55 countries
1960-77
F e d e r Cross-section
(1983)
31 countries
1964-73
Kavoussi
Cross-section
(1984)
73 countries
Jung &
Marshall
(1985)
Darratt
(1987)
Salvatore &
Hatcher
(1991)
Greenaway
& Sapsford
(1993)
Econometric Technique
Rank correlation (per capita
GNP growth on growth of
export share)
Rank correlation, Production
function, OLS (GNP growth
on export growth or real
export growth)
Linear models, OLS (change
in GDP on lagged exports)
OLS (GDP growth on
export share or export
change/output)
Production function, OLS
(GDP growth on export
growth)
OLS (GDP growth on
export growth or export
change/output)
Rank correlation, Production
function, OLS (GDP growth
on export growth)
Time series
ML, Simultaneous linear
1950-81
functions, Granger causality
37 countries
(real GNP (GDP) growth
and lagged real export
growth)
Time series
OLS white test for causality
1955-82
(real GDP growth on real
4 countries
export growth and lagged
real export growth)
Time series
Production function OLS
1963-85
(real GDP growth on real
26 countries in 4 export growth)
groups by trade
policy orientation
Cross-section
OLS tests for
104 countries
heteroskedasticity
Ave of 1960-73, (real GDP growth on
1973-90, 1980-88 growth in export ratio)
Source: Extract from Greenaway and Sapsford ((1994) 154-155).
182
Other Variables
None
Labour force growth
Domestic investment
and foreign investment
/output
Country dummies,
Direct investment,
Other foreign capital
Trade balance
Current account
Labour force growth
Investment growth
Labour force growth
Investment/output
Labour growth rate
Capital growth rate
Lagged GNP (GDP)
growth
None
Labour input growth
Capital input growth
Growth in industrial
production
None
Conclusions
Support for export
growth hypothesis
Threshold effect
Support for export
growth hypothesis
Support for export
growth hypothesis
Support for export
growth hypothesis
Support for export
growth hypothesis
Threshold effect
Support for export
growth hypothesis
Support for export
growth hypothesis
Threshold effect
Little support for
export growth
hypothesis
Rejection of export
growth hypothesis
in 3 out of 4
countries
Support for export
growth hypothesis
Support for export
growth hypothesis
Indirect evidence
of threshold effect
Primary-Export-Led Growth: The Evidence of Ghana
These studies have a major methodological problem. They correlate growth measured
by the change in the national product (whether total or per capita) with the change in exports.
Since exports are themselves part of the national product, a simultaneity problem is present,
and a positive correlation of the two variables is almost inevitable, whatever their true causal
relationship (Michaely (1977)). Although the hypothesis of export promotion clearly implies
a correlation between exports and real output growth, so do other hypotheses with quite different
policy implications. Jung and Marshall (1985) provide some examples where output growth
may not be due to any particular export promotion policies. Thus, while export growth may
lead to output growth, an equally plausible hypothesis is the reverse flow of causality - that
output growth causes export growth. Chenery and Syrquin’s cross-country study found that,
as incomes per capita rose from around US$300 to US$4,000 (in 1980 prices), the average
export share of GDP rose from 15 to 21% (Snodgrass et al. (1995)). It is therefore of interest
to study the relationship between output growth and export growth by recourse to causality
tests or analysis rather than OLS estimates. We briefly review the causality concept in
econometrics in the next sub-section.
2. Causality Concepts in Econometrics5
Feigl’s philosophical definition: According to Feigl (1953), “ The clarified (purified)
concept of causation is defined in terms of predictability according to a law (or more
adequately, according to a set of law s).” This definition, most importantly, links causation
not just to predictability but to predictability according to a law or set of laws. According
to this philosophical definition, predictability without a law or set of laws, or, as econometricians
might put it, without theory, is not causation. Hence, linking predictability to a law or set
of laws is critical in appraising various tests of causality.
Simon’s Causal Ordering and Identifiability: Simon’s (1953) notion of causality is
that “ causal orderings are simply properties of the scientist’s model, properties that are subject
to change as the model is altered to fit new observations.” Simon points out that such a
notion of causality can be applied to probabilistic as well as deterministic models. Importantly,
his notion of causality is a deductive logical concept relating to models’ characteristics, not
to empirical features of the world that require statements of inductive logic. Yet, it is the
inductive aspect of this and other models that is critical for appraising the degree to which
the model is causal (Zellner (1984)). Sims (1977) builds on Simon’s notion of causality.
The Basmann-Strotz-Wold Definition of Causality: Strotz and Wold (1960) write:
For us, however, the word (causality) in common scientific and statistical-inference usage
has the following general meaning. Z is a cause of Y if, by hypothesis, it is or would be possible
by controlling Z indirectly to control Y, at least stochastically. But it may or may not be possible
by controlling Y indirectly to control Z. A causal relation is therefore in essence asymmetric,
in that in any instance of its realization it is asymmetric. Only in the special cases may it be
reversible and asymmetric in a causal sense. These are the cases in which sometimes a controlled
change in Z may cause a change in Y and at other times a controlled change in Y may cause
5. For an extensive discussion on causality concepts and the empirical tests of causality, see Zellner (1984).
183
JOURNAL OF ECONOMIC DEVELOPMENT
a change in Z, but Y and Z cannot both be subjected to simultaneous controlled changes independently
of one another without the causal relationship between them being violated.
The Basmann-Strotz-Wold definition differs from Feigl’s by bringing in the concept of
controlled changes in variables. Note in passing, however, that the Basmann-Strotz-Wold
definition is still subsumed under Feigl’s more general definition of causation and causal laws.
The Wiener-Granger concept of causality: Granger (1969) provides an alternative theory
of causality to those of Simon, Strotz, Wold and Basmann which is based on a suggestion
by Wiener. Formally, if
is a stationary stochastic process, let
represent the set of
past values {
,
} and
}. Further let
represent the set of past and present values {
represent the set {
optimum, unbiased, least-squares predictor of
Thus, for instance,
,
}. Denote the
using the set of values
will be the optimum predictor of
since time
.
. Let
, and let
by
.
using only past
The predictive error series will be denoted by
be the variance of
,
.
. Let
be all the information in the universe accumulated
denote all this information apart from the specified series
Given these assumptions, we then have the following definition of causality: If
(1)
we say that
is causing
, denoted by
if we are better able to predict
using all available information than if the information apart from past
had been used.
Empirically, the Granger (1969) causality test is specified as follows: if we let
and
be two stationary time-series with zero means, then the simple (i.e., linear in parameters)
causal model is:
(2)
(3)
where
and
, and
are taken to be two uncorrelated white-noise series, i.e.,
for all
,
. In Equations (2) and (3),
,
,
,
and
can equal
infinity, but in practice, due to the finite length of the available data,
,
,
and
will be assumed to be finite and shorter than the given time-series. The causality tests to
be performed may be stated as:
(a) Unidirectional causality from
to
:
184
causes
if
,
Primary-Export-Led Growth: The Evidence of Ghana
can be rejected and (b) does not hold.
(b) Unidirectional causality from
to
:
causes
if
,
can be rejected and (a) does not hold.
(c) Feedback or bilateral causality is said to occur if both (a) and (b) hold.
(d) Independence is said to occur if neither (a) nor (b) holds.
To determine the causal relationship between
and
, we first fit Equation (2)
by OLS and obtain the unrestricted residual sum of squares (
). We then run another
regression equation under the null hypothesis that all coefficients of the lagged values of
are zero and obtain the restricted residual sum of squares (
we apply the
). To test the null hypothesis,
-test
(4)
which follows the
distribution with
and (
) d.f. Here,
is equal to the number
of lagged
terms,
is the number of parameters estimated in the unrestricted regression,
and
is the sample size. If the null hypothesis can be rejected, then we conclude that
the data show causality.
Several characteristics of Granger’s definition of causality may be noted. First, the use
of the series
representing all available information makes the definition nonoperational.
Granger (1969) suggests replacing “all the information in the universe” with the concept of
“all relevant information.” Second, the definition is unusual in that embedded within it is
a particular confirmatory criterion, the variance of the forecast error of an unbiased least squares
predictor. This confirmatory criterion is not applicable to processes that do not possess finite
moments. Also, for most processes, even just stationary processes, which involve parameters
whose values must be estimated from finite sets of data, an unbiased least squares “optimum”
predictor is often not available. As solutions to these problems, Granger (1969) writes:
In practice it will not usually be possible to use completely optimum predictors, unless all
sets of series are assumed to be normally distributed, since such optimum predictors may be nonlinear
in complicated ways. It seems natural to use only linear predictors and the above definitions
may again be used under this assumption of linearity.
The main problem associated with Granger’s definition is the failure to recognize explicitly
the role of economic laws or theories in defining causality (Zellner (1984)). One gets the
impression that the concept of causality can be defined entirely in terms of statistical considerations,
a point of view that is contrary to the views of Feigl, Basmann, and others. Granger’s definition
could be broadened to ameliorate these setbacks especially by the inclusion of economic laws
and/or theories. Indeed, this last approach seems very similar to that employed traditionally
in econometrics. We next review some of the empirical studies of causality.
185
JOURNAL OF ECONOMIC DEVELOPMENT
3. Empirical Studies of Causality
The empirical literature on causality is large, and we cannot do justice to it here. Some
empirical studies of causality which relate closely to this discussion include the following.
Jung and Marshall (1985) applied the Granger causality test to 37 developing countries to
determine the direction of causality between output growth and export growth in order to
test the validity of the export promotion hypothesis. The general conclusion from their study
is that the evidence in favour of export promotion is weaker than previous statistical studies
had indicated.
Serletis (1992) investigates the causality (using the Granger test) between exports, imports
and GNP growth for Canada using data from 1870 to 1985. He tests for unit roots and
cointegration of the variables and finds that GNP, export and import are not cointegrated.
His main finding is a support for the export-led growth strategy in that expansion in exports
promotes the growth of national income.
Bahmani-Oskooee et al. (1992) also found inconclusive results about the causal relations
between exports and economic growth for 20 developing countries. Five countries (Dominican
Republic, Indonesia, Korea, Taiwan and Thailand) exhibited positive causality, while three
countries (El Savador, Paraguay and Peru) indicated negative causality from exports to economic
growth. Positive causality was found from output to export growth in four countries (Korea,
Nigeria, South Africa and Thailand), while Indonesia exhibited negative causality running from
output growth to exports.
Dodaro (1993) used the Granger causality test for 87 developing countries in a study
of the export promotion hypothesis. He found no support for the neoclassical theory in any
of the NIC countries. Dodaro’s results were in favour of the export-led growth hypothesis
only in seven primarily poor and low-income countries.
Love (1994) found overall stronger support for causality from exports to economic
growth than in previous studies for 20 developing countries. Love’s result exhibited positive
and statistically significant unidirectional causality from exports to output growth in the case
of seven countries and negative causality in the case of four. Statistically significant and
bi-directional causality was found for three countries. Love interprets his results as a sign
of substantial support for the hypothesis that export growth causes growth of output.
Yaghmaian (1995) used the Granger causality test to test the causal relation between
export orientation and economic growth for a sample of 44 countries in Africa. The conclusion
from this study is that the case for export-led growth in Africa is very weak and unsubstantiated
by empirical evidence. For 38 of the countries in the sample, the causality test does not
support the contention that better export performance causes economic growth. The results
also fail to support the opposite contention that economic growth causes the growth of exports.
The review of causality tests above clearly indicates that the dominant technique is the
Wiener-Granger concept of causality. We will thus employ the Granger (1969) causality test
in this study. Note, however, that, to make the Granger (1969) concept of causality operational,
several issues must be investigated. First, one must have an economic theory, or in Feigl’s
sense, a law or set of laws. In this study, this condition is satisfied by the staple theory
of growth. Second, the variables in the analysis must be linear and stationary which calls
186
Primary-Export-Led Growth: The Evidence of Ghana
for the time-series analysis of the variables. Here, the Phillips-Perron (1988) unit root test
of stationarity will be used to test for stationarity. Third, the regression residuals must be
normally distributed. This can be tested using the Jarque-Bera (1987) test. Fourth, the issue
of mis-specification is also important. The Hausman (1978) specification test for simultaneity
and Schwarz (1978) criterion for lag lengths will be employed. The exact form of Equations
(2) and (3) will be determined using Hendry and Richard’s (1983) model specification error
test. Also, equation specification errors will be tested using the Ramsey (1969) RESET test
and residuals tests. Finally, the causality tests are carried out. These tests will be performed
in order to analyse the causality between exports and economic growth in Ghana, the subject
matter of the next section.
V. Estimation, Analysis and Conclusion
This section presents the empirical estimates for the causal relation between exports
and growth for Ghana from 1960 to 1992 (see Appendix for the data sources and notes).
The study period was determined by data availability. The variables in this analysis are the
growth rates of real gross national product (
) and real export value (
).6
1. Tests of Unit Roots and Cointegration
One major assumption of the causality test is that the time-series variables are stationary.
The stationarity properties of the variables were examined using the Phillips-Perron (1988)
test with a specification which includes a constant and a trend variable. The Phillips-Perron
(1988) tests have been widely used in testing for unit roots since they are known to have
decent power against the stationarity alternative hypothesis. The Augmented Dickey-Fuller
(1984) tests (ADF) have been popular, too. Although the ADF tests are more or less correct
under the null of a unit root, they are not very powerful (Lee and Mossi (1996)). Diebold
and Rudebusch (1991) found the ADF tests to have quite low power. Balke and Fomby
(1991) found the standard Dickey-Fuller critical values to result in too many rejections of
the unit root null hypothesis. See also Schwert (1989) and Serletis (1992) for further discussion.
The results of the Phillips-Perron unit root tests are presented in Table 4.
Table 4
Variable
RGNP
REXP
a
b
Phillips-Perron (1988) Stationarity Test Results
z-statisticb
-32.0900
-32.4940
-statistica
-5.5193
-5.6886
Conclusion
Stationary
Stationary
The asymptotic critical value at 10% of the -test is -3.13.
The asymptotic critical value at 10% of the z-test is -18.2.
From the results, we see that both the growth rate of real GNP and the growth rate
6. To get real GNP and real export value, GNP and total export value were deflated by the GDP deflator and
the export price index, respectively.
187
JOURNAL OF ECONOMIC DEVELOPMENT
of real exports are stationary since the test-statistics are greater than the 10% asymptotic critical
values (i.e., we may reject the unit root hypothesis). Hence both variables are integrated
of order zero, that is, I(0). This also implies that real GNP and real export value are integrated
of order one, that is, I(1) (as we would expect). We thus use the levels of the variables
in the analysis. A test for cointegration of the residuals of the variables indicated no long-run
relation between them. The absolute value of the test-statistic is 2.8570 which is less than
the asymptotic critical value at 10% which is 3.50 for the specification with constant and trend.
2. Model Specification and Tests
Model specification: We used the Hendry and Richard (1983) specification test (the
TTT methodology) as well as the Schwarz (1978) Criterion to determine the appropriate form
of the causality model. We selected the following form of the VAR model for the causality
tests:
(5)
(6)
Note that our specification departs from Granger’s Equations (2) and (3) in one important
way. Equations (2) and (3) are treated as single equations. However, since our discussion
makes room for simultaneity, our VAR model is specified as a simultaneous-equation system
by including current values
and
as explanatory variables in Equations (5)
and (6). This will enable us to test for simultaneity (Hausman (1978)) between the endogenous
variables. We will perform the causality tests based on the VAR model. But first we have
to evaluate the results of the regression equations. Due to space limitations, we present here
only the evaluation of the crucial Equation (5), the Export-GNP causality equation.
Simultaneity test: If the VAR model is affected by simultaneity bias, the OLS estimators
are inconsistent. We use the Hausman (1978) specification test to analyse simultaneity bias
in our model. Using the Hausman (1978) methodology, we first regress the growth rate of
real GNP (Equation (5)) on all the lagged pre-determined variables in the VAR model excluding
the current growth rate of real export value. This is the reduced-form regression. The predicted
growth rate of real GNP and the estimated residuals are substituted for the growth rate of
real GNP in Equation (6). The t-statistic of the estimated residuals is 0.88373 which is less
than the t-critical value of 2.120 for 16 d.f. at
= 0.05 and p-value = 0.3899 for a twotailed test. Since the estimated residuals are statistically insignificant, we conclude that there
is no simultaneity bias in the VAR specification.
Residuals tests: The results of the residuals tests (Normality, Equation specification error,
ARCH(p)) are summarized in Table 5 below:
188
Primary-Export-Led Growth: The Evidence of Ghana
Table 5
Normality, Equation Specification Error and ARCH(p) Tests
TestDeg.
Critical
Type of Test
P-value
Conclusion
statistic of freedom
value 5%
NORMALITY
Approximately normally
JARQUE-BERA(JB)
0.1337
2
0.940
distributed
RAMSEY RESET
RESET(2)
4.1349
F(1, 6)
5.99
No mis-specification bias
RESET(3)
4.2620
F(2, 5)
5.79
No mis-specification bias
RESET(4)
5.1252
F(3, 4)
6.59
No mis-specification bias
ARCH(p)
ARCH(1)
0.5260
0.468
No autoregressive
ARCH(2)
0.6660
0.625
c o n d i t i o n a l
ARCH(3)
0.7743
0.825
heteroscedasticity
The normality test indicates that the p-value exceeds 5%, and we may conclude that
the residuals of the regression equation are approximately normally distributed. We see that
the RESET tests are statistically insignificant at the 5% level of significance, which indicates
no equation specification error. With time-series data, it is pertinent to test for the twin problems
of autocorrelation and heteroscedasticity. This is done using the autoregressive conditional
heteroscedasticity (ARCH(p)) model.7 The results indicate that the error variance is not affected
by autoregressive conditional heteroscedasticity.
Test for multicollinearity: Multicollinearity is checked by using the Tolerance (
)
and the Variance Inflation Factor (
) tests from the auxiliary regressions.8 Since the
values (falling between 1.700 and 2.434) are all less than 10, and the
values (falling
between 0.411 and 0.617) are not very close to zero, we may conclude that the specification
is not seriously affected by multicollinearity. Since even with near collinearity the OLS
estimators are still BLUE and the problem of multicollinearity is a question of degree rather
than the absence of multicollinearity, we may accept the specification for our purpose.
3. The Causality Tests
The results of the Granger (1969) test of causality between the growth rate of real GNP
and real export value are presented in Table 6.
Table 6
Direction of Causality
Results of the Granger (1969) Causality Tests
F-statistic
F-critical at
REXP
RGNP
30.18
3.59
RGNP
REXP
3.00
3.72
=0.01
Decision
Reject
;
that is, there is causality
Do not reject
;
that is, there is no causality
7. The ARCH(p) model has the test statistic
where n = number of observations and
= coefficient of
determination from the auxiliary regression of the square of the residuals on its different lags (see R. Engle
(1982)).
8. The VIF j is defined as
remaining
, where
is the
in the (auxiliary) regression of
on the
regressors; Tolerance is also defined as TOL j = 1/VIF j (see D.G. Kleinbaum et al. (1988)).
189
JOURNAL OF ECONOMIC DEVELOPMENT
The results indicate that the growth rate of real export value do cause (in Granger-sense)
the growth rate of real GNP for the study period. The reverse causality test (from GNP
to Exports) does not hold. Thus, in the first case, the null hypothesis may be rejected, indicating
unidirectional causality between Exports and GNP. On the other hand, the growth rate of
real GNP does not cause the growth rate of exports.
The causality tests indicate that Ghana has been able to spread the dynamic gains from
trade to other sectors of the economy. For the period under study, the support for the benefits
of economies of trade and positive externalities (the dynamic gains) does exist in Ghana.
In other words, primary exports have been a major growth-generating force during the study
period. We may postulate that Ghana has gained from its comparative advantage in primary
commodities as a source of economic growth. The positive role of the primary sector in
the economic growth process is thus comparable to that of Indonesia, Malaysia and Thailand.
Bahmani-Oskooee et al. (1992) found positive causality between exports and economic growth
for Indonesia and Thailand. Our findings reinforce that of Serletis (1992) who uses similar
methods and finds that the evidence in favour of export promotion is stronger than previous
statistical studies have indicated.
4. Conclusion
This paper has attempted to test the hypothesis of primary-export-led growth using Ghana
as a case study for the period 1960-1992. The theoretical framework used in the study is
the staple theory of growth. A staple is a primary product that faces a large and growing
demand in world markets, does not require elaborate processing, and has a high enough value-toweight (volume) ratio to bear transportation costs. The staple theory emphasises the fact that
staple exports are the leading sector of the economy which set the pace for economic growth.
Economic growth is a consequence of the trickle-down effects of the primary export sector.
The inducement to domestic investment resulting from the increased activity of the export
sector can be disaggregated into three linkage effects: backward-, forward- and final
demand-linkage. These linkage effects are analogous to the dynamic gains from trade (see
the introduction) as stressed by neoclassical economic theory.
Several econometric tests were performed, the ultimate test being the Granger (1969)
causality test. The test indicated unidirectional causality between the growth rate of real
GNP and real Exports. In other words, export growth is a cause (in Granger-sense) of the
growth of real GNP (or economic growth). Thus Ghana has been able to spread the dynamic
gains from trade to other sectors of the economy. For the period under study, the support
for the benefits of economies of trade and positive externalities (the dynamic gains) exists
in Ghana. In other words, primary exports have been a major growth-generating force during
the study period. We may postulate that Ghana has gained from its comparative advantage
in primary commodities as a source of economic growth. Our findings reinforce those of
Love (1994) and Serletis (1992) that the evidence in favour of export promotion is stronger
than previous statistical studies have indicated.
One concern is related to the data used in the analysis. Given the limited data used
in the study, the evidence is suggestive and may not be conclusive. Again, since the data
190
Primary-Export-Led Growth: The Evidence of Ghana
are for one country, the results may have limited implications. However, the results might
have greater reliability since the model has avoided the problems of cross-country analysis.
It may also be of interest to replicate this study using data from other developing countries
to assess the validity of the export promotion hypothesis.
191
JOURNAL OF ECONOMIC DEVELOPMENT
Appendix
Data Sources and Notes
Gross National Product (GNP) in million cedis at current prices - Statistical Service, Accra:
Quarterly Digest of Statistics, various issues; Economic Survey, various issues ; 1990-1992
from World Bank: World Tables, 1994 issue. World and country GNP data are from UN:
National Accounts Statistics International Tables and World Bank: World Tables , various
issues.
Export (EXP) value in million cedis at current prices - Statistical Service, Accra: Quarterly
Digest of Statistics, various issues; External Trade Statistics, various issues; 1990-1992 from
World Bank: World Tables, 1994 issue. The data were in US$m and were converted to cedis
using the annual average cedi/$ exchange rates. Data on percentage distribution of exports
for Ghana are from Huq ((1989) Table 10.5). Data on primary exports as a percentage of
world exports are from the UN: Yearbook of International Trade Statistics, various issues.
Export Price Index (EPI) - IMF: International Financial Statistics, various issues; World
Bank: World Tables, 1987 and 1994 issues. The data were in 1958, 1963, 1980 and 1987
base years and were converted to 1980 common base year.
Nominal GDP (NGDP) and Real GDP (RGDP) in million cedis - 1960-1984 from M.M.
Huq ((1989) Table A.1, p. 287); 1985-1992 from IMF: International Financial Statistics, various
issues. The real GDP data were in 1970 and 1985 base years and were converted to 1970
common base year. The GDP deflator is estimated using the following formula: GDP deflator
= [(N GDP/RGDP)× 100].
Land and Labour - FAO: Production Yearbook Vol. 30 (1976) and Vol. 45 (1991). Land
area refers to total area, excluding area under inland water bodies. Labour is the economically
active population.
192
Primary-Export-Led Growth: The Evidence of Ghana
References
Bahmani-Oskooee, M. et al. (1991), “Exports, Growth and Causality in LDCs: A Re-examination,”
Journal of Development Economics , Vol. 36, Amsterdam, 405-415.
Balassa, B. (1965), “Trade Liberalization and Revealed Comparative Advantage,” Manchester
School of Economic and Social Studies, Vol. 33(2), Manchester, 99-124.
Balke, N.D., and T.B. Fomby (1991), “In frequent Permanent Shocks and the Finite-sample
Performance of Unit Root Tests,” Economics Letters, Vol. 36, North-Holland, 269-273.
Bruno, M. (1967), “The Optimal Selection of Export-promoting and Import-substituting Projects,”
in United Nations Planning the External Sector: Techniques, Problems and Policies,
New York.
Corden, M.W. (1971), “The Effects of Trade on the Rate of Growth,” in Trade, Balance
of Payments and Growth, eds. by J.N. Bhagwati et al., Amsterdam: North-Holland,
117-143.
Diebold, F.X., and G.D. Rudebusch (1991), “On the Power of Dickey-Fuller Tests against
Fractional Alternatives,” Economics Letters, Vol. 35, North-Holland, 155-160.
Dodaro, S. (1993), “Exports and Growth: A Reconsideration of Causality,” The Journal of
Developing Areas, Vol. 27, Macomb-Illinois, 227-244.
Engle, R. (1982), “Autoregressive Conditional Heteroscedasticity with Estimates of the Variance
of United Kingdom Inflation,” Econometrica, Vol. 50, Evanston-Illinois, 987-1007.
Granger, C.W.J. (1969), “Investigating Causal Relations by Econometric Models and
Cross-Spectral Methods,” Econometrica, Vol. 37, Evanston-Illinois, 424-438.
Grant, H.M., and M.H. Watkins (1993), Canadian Economic History: Classic and Contemporary
Approaches , Ottawa: Carlton Univ. Press Inc., 19-38.
Greenaway, D., and D. Sapsford (1994), “What does Liberalization do for Exports and Growth?”
Weltwirtschaftliches Archiv, Vol. 130, Tubingen-Kiel, 153-174.
Hausman, J.A. (1978), “Specification Tests in Econometrics,” Econometrica, Vol. 46(6),
1251-1271.
Hendry, D.F., and J.F. Richard (1983), “The Econometric Analysis of Economic Time Series,”
International Statistical Review, Vol. 51, Harlow, 3-33.
Huq, M.M. (1989), The Economy of Ghana: the first 25 years since Independence, London:
MacMillan Press.
Jarque, C.M., and A.K. Bera (1987), “A Test for Normality of Observations and Regression
R esiduals,” International Statistical Review, Vol. 55, Harlow, 163-172.
Jung, W.S., and P.J. Marshall (1985), “Exports, Growth and Causality in Developing Countries,”
Journal of Development Economics , Vol. 18, Amsterdam, 1-12.
Kapur, Ishan et al. (1991), Ghana: Adjustment and Growth, Washington, D.C.: IMF.
Kleinbaum, D.G. et al. (1988), Applied Regression Analysis and other Multivariate Methods,
PWS-Kent, 2nd ed., Boston, 210.
Leamer, E.E. (1984), Sources of Comparative Advantage, Cambridge: MIT.
Lee, J., and D. Mossi (1996), “On Improvements of Phillips-Perron Unit Root Tests Using
Optimal Bandwidth Estimates,” Applied Economics Letters, Vol. 3, London, 197-200.
193
JOURNAL OF ECONOMIC DEVELOPMENT
Love, J. (1994), “Engines of Growth - The Exports and Government Sectors,” World Economy,
April, 203-218.
Michaely, M. (1977), “Exports and Growth: An Empirical Investigation,” Journal of Development
Economics, Vol. 4, Amsterdam, 49-53.
Phillips, P.C.B., and P. Perron (1988), “Testing for a Unit Root in Time Series Regression,”
Biometrika, Vol. 75(2), Cambridge, 335-346.
Ramsey, J.B. (1969), “Tests for Specification Errors in Classical Linear Least Squares
Regression Analysis,” Journal of the Royal Statistical Society , Series B, Vol. 31,
London, 350-371.
Roemer, M. (1970), Fishing for Growth: Export-Led Development in Peru, 1950-1967,
Cambridge-Mass.: Harvard Univ. Press, 5-23.
Sachdev, S. (1993), “International Competitiveness and Agricultural Export of India,” Indian
Economic Review, Vol. XXVIII, No. 2, Delhi, 203-217.
Schwarz, G. (1978), “Estimating the Dimension of a Model,” The Annals of Statistics, Vol.
6(2), California: Hayward, 461-464.
Schwert, G.W. (1989), “Testing for Unit Roots: a Monte Carlo investigation,” Journal
of Business and Economic Statistics, Vol. 7, Washington, D.C., 147-159.
Serletis, A. (1992), “Export growth and Canadian Economic Development,” Journal of
Development Economics, Vol. 38, Amsterdam, 133-145.
Snodgrass, D.R. et al. (1995), Economics of Development, 4th ed., W.W. Norton & Co., New
York.
White, K.J. (1993), Shazam The Econometrics Computer Program Users Reference Manual,
Version 7.0, Vancouver: McGraw-Hill Book Co.
Yaghmaian, B. (1995), “Exports and Economic Growth in Africa: a Note on the Causality
Question,” Review of Radical Political Economics, Vol. 27(3), Cambridge-Mass., 72-85.
Zellner, A. (1984), Basic Issues in Econometrics , Chicago: The Univ. of Chicago Press, 35-74.
194